bluelogo.gif (1643 bytes)

home page
people
research
publications
seminars
travel
search
UCL

Products of Hidden Markov Models

Andrew Brown
Department of Computer Science
University of Toronto, Canada

and

Geoffrey Hinton
Gatsby Computational Neuroscience Unit
University College London
London

T. Jaakkola and T. Richardson eds., Proceedings of Artificial Intelligence and Statistics 2001, Morgan Kaufmann, pp3-11

Abstract
We present products of hidden Markov models (PoHMM's), a way of combining HMM's to form a distributed state time series model.  Inference in a PoHMM is tractable and efficient.  Learning of the parameters, although intractable, can be effectively done using the Product of Experts Learning rule.  The distributed state helps the model to explain data which has multiple causes, and the fact that each model need only explain part of the data means a PoHMM can capture longer range structure than an HMM is capable of.  We show some results on modelling character strings, a simple language task and the symbolic family trees problem, which highlight these advantages.


Download:  [ps] [ps.gz] [pdf]